Influence of hot-carrier luminescence from avalanche photodiodes on time-correlated photon detection

advertisement
758
OPTICS LETTERS / Vol. 25, No. 10 / May 15, 2000
Influence of hot-carrier luminescence from avalanche
photodiodes on time-correlated photon detection
Gökhan Ulu
Department of Physics, Boston University, 590 Commonwealth Avenue, Boston, Massachusetts 02215
A. V. Sergienko and M. S. Ünlü
Department of Electrical and Computer Engineering, and Photonics Center, Boston University,
8 Saint Mary’s Street, Boston, Massachusetts 02215
Received December 6, 1999
We present the results of our time-resolved measurements of hot-carrier luminescence from passively quenched
Geiger-mode avalanche photodiodes. In time-correlated photon-counting (TCPC) experiments, hot-carrier
luminescence interferes overwhelmingly with the coincidence spectrum, which results in artifacts. This
potential problem should be taken into account in setting up TCPC experiments.  2000 Optical Society
of America
OCIS codes: 030.5260, 230.5170, 130.0250, 190.4410.
Single-photon avalanche photodiodes (APD’s) have
been established as a strong alternative to photomultiplier tubes in photon-counting applications because
they provide comparable performance, are small
and rugged, and require simpler and less-expensive
electronics.1 Single-photon sensitivity is achieved by
reverse biasing of the diode above breakdown. The
absorption of a photon triggers an avalanche of hot
carriers, which results in a large-current pulse. Hotcarrier luminescence in reverse-biased p –n junctions
is a known concept in the literature.2,3 We observed
the invasive effects of hot-carrier luminescence from
avalanche photodiodes (APD’s) in a time-correlated
photon-counting experiment. We operated commercially available silicon APD’s (EG&G, C30902S-TC) in
Geiger mode to record coincidences between entangled
photons.4 Extraneous peaks were recorded in the coincidence spectrum, overwhelming the data of interest.
In this Letter we report a preliminary examination
of the nature of this observation and its inf luence
on time-correlated photon-counting experiments that
utilize APD’s.
The essence of the experiment (the setup is shown
in Fig. 1) is to observe quantum correlation between
entangled photon pairs produced from the laser pump
by parametric downconversion process in a nonlinear crystal (LiIO3 ). The detectors are reachthrough APD’s with active diameters of 0.5 mm and
are cooled by double Peltier stages. They are connected to the start and stop inputs of a standard
time-correlated photon registration setup via simple
passive-quenching circuits.1 During the experiments,
we encounter coincidence spectra such as that plotted
in Fig. 2. The peaks extend over a temporal region of
⬃100 ns. The spectrum persists with the pump source
turned off and with a mirror replacing the crystal facet,
excluding the possibility of a peculiar phenomenon in
the crystal or of stray light owing to the pump. These
symptoms indicate the likely origin of this artifactual
spectrum to be the avalanche emission from each
APD detected by the other. This claim is consistent
with the already existing literature on hot-carrier
0146-9592/00/100758-03$15.00/0
visible –IR luminescence in reverse-biased p –n junctions. The peaks on the left and the right are due to
the emission events from the stop and the start detectors, respectively. The avalanches in the detectors
can occur as a result of the detection of any photon
in the surroundings or of generation of a random
carrier in the absorption region (dark noise). Each
coincidence event is a result of registering the time
difference between the occurrence of the current pulse
that is due to the avalanche in one of the detectors
and the detection of a photon, emitted during that
avalanche, by the other detector. On the left peak,
associated with the stop detector, the decay appears
at earlier times because the detection of the emitted
photons initiates the coincidence event, whereas, for
the right peak, the avalanche current pulse triggers
the coincidence circuit.
Fig. 1. Setup for time-correlated measurement of entangled photons. The squares show different possibilities for
overlap of APD imaging paths. The shaded circles are
the entangled photon streams, and the larger circles are
regions imaged by the APD’s: 1, no overlap; 2, overlap
owing to misalignment; 3, inevitable overlap at optimum
alignment. TAC, time-to-amplitude converter; MCA, multichannel analyzer.
 2000 Optical Society of America
May 15, 2000 / Vol. 25, No. 10 / OPTICS LETTERS
Fig. 2. Sample luminescence decay spectrum. We turn
off the main experiment by shutting down the pump to
emphasize the interfering phenomenon.
An explanation of our observations requires careful
consideration of the optical conf iguration of the setup.
The entangled photons, conventionally named the signal and the idler, emerge from the crystal at angles
given by phase-matching conditions according to their
frequency. The detectors and the optical components
that are placed in the path of signal and idler photon
streams image two regions on the crystal exit surface.
When both signal and idler branches are at small and
nearly equal angles to the facet normal, optimum alignment may lead to the existence of an optical path between detectors A (start) and B (stop). The path starts
from the window of one detector, passes through the
components in one branch, ref lects off the crystal facet
(region F), and, passing through the other branch, ends
on the other detector. We have observed that the existence of the coincidence spectrum depends on the existence of this optical path, A –F–B. The strength of
the optical coupling between the two detectors depends
on the overlap of the imaging volumes of the APD’s
on the crystal surface either required by the optimal
positional settings of the optical components or as a result of to their misalignment. Because the detection
is at single-photon-counting level, even weak optical
coupling will cause luminescence interference in the
experiments. Its amount and temporal range make
luminescence a prohibitive problem, especially when
the signal of interest is weak. The optical path between the detectors cannot be spatially isolated merely
by alignment in experiments in which photon pairs are
created collinearly.
We also studied the bias voltage and temperature dependence of the time-resolved luminescence. Several
mechanisms have been proposed for describing light
emission from reverse-biased Si p –n junctions based
on spectral observations.3,5,6 Lacaita et al.7 reported
that photon emissions are responsible for spreading
the avalanche in large area reach-through APD’s such
as the ones used here. They pointed out that photonassisted avalanche growth sets the ultimate limit to
the timing performance of the APD’s. Preliminary
examination indicates that the most of the observed
luminescence energy lies in the range 1.1–2.0 eV (650–
1100 nm), which is consistent with indirect interband
759
transitions. The number of photons detected by the
stop detector is 1–10 photons per avalanche event on
the start detector. Taking into account the transmission of the optical elements and apertures and the
detection eff iciency of the stop detector, we estimate
that the number of photons emitted per avalanche into
2p sr is of the order of 103 104 . A typical avalanche
current pulse lasts approximately 10–20 ns, with a
peak at a few milliamperes, resulting in 107 108 carriers crossing the junction. Therefore the estimated
eff iciency of the hot-carrier emission is approximately
1025 , which is consistent with previous reports.7
The luminescence events observed in the coincidence
data are not simple optical pump-decay processes.
The temporal behavior of emissions is expected to
be a convolution of the avalanche process, which is
the major driving mechanism of the hot-carrier population, with optical relaxation mechanisms. Because
these mechanisms are much faster than the multiplication process, the observed spectrum is approximately
the time evolution of the charge population during
the avalanche. The inset in Fig. 3 shows the timeresolved emission peak when all wavelengths are collected. The data are fitted to a biexponential with
time constants t1 and t2 . Although the temporal response of the receiving (stop) detector is not deconvolved from the data, it remained constant throughout
the measurements. As shown in Fig. 3, t1 and t2 decrease when the bias voltage is increased. The decay
side of the peak is similar to the oscilloscope trace of
the output pulse of the APD and shows the same dependence on bias as t1 and t2 . However, t1 and t2 remain
fairly constant, with an ambient chip temperature that
varies from 225 to 25 ±C. The coincidence spectrum
persists in a large time interval for any practical combination of bias voltage and temperature.
The obvious solution to the problem of reducing the
inf luence of APD emissions is to make sure by design
that no direct optical path exists between the two detectors as a result of specular ref lection off an intermediate surface such as the nonlinear crystal in our
experimental case. So we chose the cut angle 共u兲 of our
thick nonlinear crystal such that the entangled photon streams of interest exit the surface through wellseparated sections. The cut angle for thin crystals
Fig. 3. Voltage dependence of the luminescence decay
constants.
760
OPTICS LETTERS / Vol. 25, No. 10 / May 15, 2000
can be chosen appropriately to work with the pump
at off-normal incidence so the normal of the crystal
surface does not coincide with the bisector of the entangled photon streams. Additionally, one can utilize
bandpass filters if the optical spectrum of the signal
of interest and the APD emission do not overlap. Another solution arises when we note that the APD’s were
passively quenched: In several other similar experimental conf igurations with actively quenched APD’s,
no luminescence-related artifacts were noticed, even
though the photon pairs were collinear and the optomechanical alignment required full overlap of optical
paths leading to the APD’s. In active quenching, the
avalanche is suppressed by an active circuit. Shortly
after the onset of the avalanche, the circuit lowers the
bias voltage rapidly below breakdown and halts the
growth of the avalanche process long before it can reach
its passively quenched limit. However, depending on
the settings, the avalanche proceeds passively for
5–15 ns. For the same detector, actively quenched operation will produce a smaller population of hot electrons than passively quenched operation if the timing
parameters of the circuit are shorter than the stray and
junction capacitance-charging times of the detector.1
Another possible side effect of luminescence is an
interesting external afterpulsing effect. The photons
emitted by the avalanche of a registered photon may
ref lect off a normal surface in the imaging path back
onto the detector itself. The ref lected photons will
keep the timing circuit busy blocking the detection
of the events of interest. Advanced active-quenching
circuits provide a hold-off time during which the
diode is kept quiescent. The hold-off feature serves to
minimize afterpulsing that is due to charge trapping.
It can actually be extended to cover the external
afterpulsing as long as the major surface of ref lection
is known and localized and its distance to the detector
is of reasonable length. In general, it is advisable to
minimize ref lective surfaces by using antiref lective
coatings where possible.
In conclusion, we have presented a report of the
effects of hot-carrier luminescence in APD’s on
time-correlated photon-counting experiments. The
artifacts observed have a direct inf luence on the timing performance of individual APD’s as well as on the
applicability of single and multiple APD’s in certain
conf igurations.
This study was supported by the U.S. Off ice of Naval
Research (N00014-96-1-0652) and the National Science
Foundation (INT96-01770).
References
1. S. Cova, M. Ghioni, A. Lacaita, C. Samori, and F. Zappa,
Appl. Opt. 35, 1956 (1996).
2. R. Newman, Phys. Rev. 100, 700 (1955).
3. J. Bude, N. Sano, and A. Yoshii, Phys. Rev. B 45, 5848
(1992).
4. L. Mandel and E. Wolf, Optical Coherence and Quantum Optics (Cambridge U. Press, Cambridge, 1995),
Chap. 22, pp. 1074 –1088.
5. N. Akil, S. E. Kerns, D. V. Kerns, Jr., A. Hoffmann, and
J-P. Charles, IEEE Trans. Electron. Devices 46, 1022
(1999).
6. D. K. Gautam, W. S. Khokle, and K. B. Garg, Solid-State
Electron. 31, 219 (1988).
7. A. Lacaita, S. Cova, A. Spinelli, and F. Zappa, Appl.
Phys. Lett. 62, 606 (1993).
Download